Re: strange umass/scsi behaviour

On Thu, 31 Jul 2003, John Hay wrote:
> > > > > Does anyone have an idea why the umass/scsi code behave differently
> > > > > between if you boot with a device already plugged in as opposed to
> > > > > plugging it in later? In my case it is a Sandisk Cruiser. If I plug
> > > > > it in before booting, it works just great, but if I plug it in later,
> > > > > it does not want to work.
> > > The
> > > "problem" is that only da0 pitch up in the /dev directory. There should
> > > be a da0s1 too. Using fdisk on da0 does show what looks like a valid
> > > partition table with one DOS partition. I have tried camcontrol reset,
> > > inquiry, start and rescan, but can't get da0s1 to make its appearance.
> >
> > Sounds like a GEOM problem.  Unless you have other errors from da0 after
> > the last line you posted, it appears the command retry attempt succeeded.
> > At that point, the problem is likely with GEOM not parsing the partition
> > table when it appears and making the slice show up.
> >
> > But I don't know enough about GEOM to claim this its fault.
>
> Ok, but why does geom pick it up correctly when booting with the disk
> plugged in? Does geom mabe get called before the disk is ready? If
> that "Opened disk ..." message is from geom, the order of things looks
> like geom gets called just after the "Unretryable error".

When you boot with it attached, USB is turned on early enough in the boot
cycle that the delay while probing SCSI/ATA devices gives the USB device
enough time to warm up before GEOM does its "taste" of the device (last in
the boot cycle).

I'm sending a copy of this to phk_at_ to see if he has any further thoughts.

-Nate